A false accept rate of 0.01% sounds like nothing, and a false reject rate of 1% sounds small, until a reader at a busy door turns them into people. This calculator does that sum. Enter the users, the entries each makes a day and the kind of reader, and choose whether it checks one template after a card or PIN (1:1) or searches the whole database (1:N). It shows how many people are refused a day and the chance that a stranger is let in, per try, a day and a year. Read fingerprint access control for the reader itself.
The sums, plainly
People refused a day = users x entries each x the false reject rate (FRR). The chance a stranger's single try is accepted is the false accept rate (FAR) when the reader checks one template. When it searches all templates, the chance rises roughly with their number: FAR x templates, a simple approximation that holds while the result is small. Over a day and a year, the tool adds up the tries strangers make. The defaults come from independent tests by a US standards institute for finger, face and iris, and a maker's datasheet for palm; type your own reader's figures over them.
Why real doors do worse
The test figures are measured on good images with careful matching. At a door, fingers are dry, dusty or worn, faces are backlit or under hard hats, and people hurry. A cheap sensor or a badly set threshold can be many times worse than the lab.
The installer sets the threshold: raise it and fewer strangers get in but more staff are refused; lower it and the reverse. The calculator shows the trade-off in people, so you can decide.
If this, then that
- If many staff are refused: enrol two or three fingers each, clean the sensor, or add a card fallback.
- If the database is large: use 1:1 with a card or PIN first, so the reader checks only one template.
- If the door guards drugs, cash or firearms: combine two factors, a card and a finger.
- If staff refuse to enrol: offer a card or PIN; under POPIA, see biometric data and POPIA.
Using the result with an installer
Ask each supplier for the reader's FAR and FRR, and how they were measured: by whom, on how many people, at which threshold. A figure with no test behind it is marketing. Ask what the reader does when someone fails three times, and how the threshold can be changed later. Put the door count and users on the request below for quotes. For the parts and power, use the access control planner; for the overview, biometric access control.
Sources
- US standards institute, proprietary fingerprint template evaluation: one leading matcher on the MINEX III data, single index finger: FNMR 0.0016 at FMR 0.0001, 0.0015 at 0.001; two index fingers fused: FNMR 0.000006 at FMR 0.0001
- Biometrics trade news, report on the US face verification test: the most accurate algorithms on the US standards institute's 1:1 test had an FNMR of 0.0025 to 0.0035 at an FMR of 0.000001 (visa photos) and 0.00001 (mugshots), 2020
- US standards institute, iris recognition evaluation report: IREX IX part one: the most accurate one to one matcher had an FNMR of 0.0057 (about 1 in 175) at an FMR of 0.00001, two eye comparisons on field collected images; four others between 0.0066 and 0.0070
- Access control maker, palm and face terminal datasheet: 5,000 palm templates, 10,000 faces, 20,000 cards, 200,000 transactions; palm authentication under 0.35 s, face under 1 s; palm FAR 0.01%, FRR 1.3%; face FAR 0.01% or less, FRR 0.02% or less; 12 V DC 3 A supply
Frequently asked questions
What is the difference between FAR and FRR?
Why does 1:N matching raise the chance of a false accept?
Which biometric has the lowest error rates?
Can someone help me choose a reader?
More calculators and tools
- Access control plannerReaders, controllers, locks, exits, power and cable for your doors
- Maglock calculatorThe holding force class, bracket, current and supply for one door
- Backup battery calculatorThe battery in amp hours to carry readers and locks through outages
- Gate motor calculatorMotor class by gate mass and width, and the cycles a battery gives
- Clocking terminal calculatorHow many clocking terminals clear the shift change in time
- Turnstile lane calculatorLanes of tripod, speed gate or full height turnstile for your peak
- Electric fence calculatorWire, brackets, warning signs and earth spikes for your fence
